Tomographic Image Reconstruction Using Forward Projection Model Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing tomosynthesis imaging techniques face challenges in accurately correcting object movement due to positional deviations between projection images, particularly because they fail to account for information from structures other than feature points.
Innovation Solution
An image processing device and method that utilize a forward projection model with optimized parameters, including absorption coefficients and object movement, to generate accurate tomographic images by aligning pixel values from pseudo-projection images with actual projection images, using a three-dimensional model with voxels and intersection lengths to correct object movement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If feature point-based positional deviation correction is used, then correction process is simple, but correction accuracy is insufficient due to ignoring other structural information
Solution Approach 1:
The patent applies universality by utilizing all projection images for dual purposes: both for generating the tomographic image and for calculating positional deviation. Instead of relying solely on feature points, the system uses the entire structural information from multiple projection images to simultaneously achieve image reconstruction and movement correction, making the correction process more comprehensive and accurate while maintaining operational efficiency through integrated processing.
Solution Approach 2:
The patent implements feedback by using the calculated positional deviation to iteratively adjust and correct the tomographic image generation process. The system calculates positional deviation from projection images, uses this information to correct image alignment, and repeats the process until convergence, ensuring high correction accuracy while maintaining computational efficiency through iterative optimization.
2Measurement precision
If all projection image information is used for correction, then correction accuracy improves, but processing complexity increases
Solution Approach 1:
The patent merges the tomographic image generation process with the positional deviation calculation process into a single integrated workflow. By combining these functions, the system uses the same projection image data for both purposes, eliminating the need for separate processing steps and reducing overall computational complexity while maintaining high correction accuracy through comprehensive use of structural information.
Solution Approach 2:
The system performs self-service by automatically calculating positional deviation from the projection images themselves without requiring external reference data or manual intervention. The projection images contain all necessary information for both tomographic reconstruction and movement correction, allowing the system to self-correct using its own input data, thereby simplifying the processing pipeline while achieving accurate results.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enables high-accuracy correction of object movement, resulting in improved image quality and reduced blurriness in tomographic images by effectively accounting for all structural information during the imaging process.
Implementation Method 1
irradiating an object with radiation emitted from a radiation source... an absorption coefficient assigned to each voxel... where a path of the radiation emitted at the irradiation position intersects the three-dimensional model
Data Source
AI summary
A CPU of a console acquires a plurality of projection images, performs an optimization process on a forward projection model, which has, as parameters, an absorption coefficient assigned to each voxel of a three-dimensional model that is virtually set in a three-dimensional space in which the object is disposed and has a plurality of voxels as constituent units, an intersection length of each voxel where a path of radiation emitted at an irradiation position intersects the three-dimensional model, and an amount of movement of the object, on the basis of the projection images at each of a plurality of irradiation positions, and generates a tomographic image of the object using the optimized parameters.